Anchored Relative Strength
I will preface this post by saying that this is a concept that I have not yet had a chance to test out. That said, I usually start first with a theory or a logical observation and proceed to creating a quantitative method to capture that insight. The concept relates to everyone’s favorite topic–relative strength which is the method of finding the best stock or asset to hold at a given time relative to the universe of alternatives. One of the typical features of relative strength investing regardless of the ranking metric used is some form of a rolling lookback window. In most of the research this tends to span from 3-12 months which coincides with the typical periodicity that most markets tends to trend. It is obvious from my own research that while relative strength works well on all lookbacks, it works a lot better on some than others. Furthermore the best lookback will vary considerably by market or asset grouping– currencies behave differently than equities as but one of many examples.
Finding the optimal lookback length on an adaptive basis is a worthwhile idea that is not the subject of this post. What is more interesting is whether the optimal lookback is a function of a persistent cycle length or whether the optimal lookback is simply an artifact of predictable fractal behavior. In other words, do the onset of significant rallies or drawdowns behave as a catalyst for change in the relative strength cycle? It is plausible for example that a significant drawdown in a bull market would “reset the clock” for industry group relative strength with new leaders replacing the old ones in such circumstances. It stands to reason that any significant drawdown or rally can 1) change risk-seeking/risk-avoiding behavior 2) can be caused by perceptions of economic health that will favor certain sectors over others 3) ignite or deflate critical volume liquidity for a given issue that will tend to reinforce trends in both directions.
If this is true, then having fixed rolling lookbacks such as 3 months or 9 months will be less effective than creating an “anchored” relative strength measure that works off of a fixed point of origin. For example, one could measure the relative strength from the bottom point of a significant drawdown after it has re-established new highs as the basis for identifying the strongest groups. Furthermore, one can also measure the relative strength from the breakout point in the previous situation. Alternatively, one could measure the relative strength from the top of a significant rally after it establishes new lows and also at the breakout point as well. One could also measure relative strength from the bottom of a major rally (ie like the march low in 2009). There are numerous possibilities using this angle, and the key is to compare such effectiveness of such methods to rolling lookbacks so that it can be determined whether they add statistically significant value over a simpler appoach. My theory is that while anchored relative strength may not consistently beat rolling relative strength, there may be some event patterns (such as those mentioned above) that do have relevance. Furthermore, this may be especially true in volatile markets versus consistently trending markets. Of course, the opposite could be true as well. It is up to the diligent system developer to explore…………